<emsyr>Hello everyone. Is there a way to skip check phase of a derivation when executing "sudo guix system reconfigure /etc/config.scm" ? When using "guix install [derivation]" I found that --without-tests=[derivation] does the work but it is not supported in "guix system"
<mhj[m]>Still learning guile, still loving guix. Right now I am working on creating a snapshot bash script for btrfs that works well with guix. When I learn more guile, I'll make a custom service for btrfs. I use ZFS with NixOS on my laptop, and it works well, so I just wanna replicate that experience on my desktop with guix :)
<civodul>daphnis: right, the "devel" manual i linked to is for the latest and greatest; you would need to run 'guix pull' first
<fi3>Hi! I created very simple rust package (just print something, just one binary crate, no dependency). Then I create a file.tar.gz with Cargo.toml and src/main.rs. Then I defined the most simple guix package to build.
<roptat>je suis pas contre une relecture non plus, je fais régulièrement des fautes d'accord
<roptat>je vais sans doute un peu trop vite quand je traduis
<apteryx>iyzsong: I think you had experimented with sitecustomize and GUIX_PYTHONPATH in the past, reading old mail threads. The idea was sound; it seems to work well! You may want to review/try the branch cu/farewell-to-pythonpath and post comments on https://issues.guix.info/46028 if you have any!
<ngz>Il faut que j'apprenne comment utiliser Weblate + Emacs…
<civodul>roptat: we should add a prominent link on the web pages of the manual to add new translations or fix existing ones
<apteryx>roptat: thanks for all this work; it's awesome!
<roptat>but there is an API, so that might be possible
<roptat>at least, you can download and upload po files
<roptat>civodul, I pushed the change to maintenance, could you reconfigure berlin when you have time?
<dongcarl>Python packagers: I'm trying to package a library which tries to call `find_library` at import time in order to locate libssl and libcrypto, is there an elegant way to make this work or is the only way just to hardcode it with a substitute* in my package definition?
<mdevos>mzbm: happens to me as well, but with gdm and particularily when running intensive ‘guix install’ and ‘guix system reconfigure’
<mzbm>By discrete you mean a dedicated graphics card?
<mzbm>Its an Nvidia unfortunately, but it did work quite well in OpenSUSE and Arch Linux
<argylelabcoat>I've got a question about guix containers. Is it possible to access a serial port? For example, if I try passing ttyACM0 into a guix container, I get this error: `guix environment: error: mount: mount "/dev/ttyACM0"`
<mzbm>mdevos: for me as well. Especially when loading websites in the browser.
<daphnis>i can't find any gnat package (needed for coreboot)
<mdevos>mzbm: I often load websites at the same time as when building stuff with guix. Maybe I'm building guix's building processes too much
<mdevos>mzbm: nvidia requires propietary firmware or something, or so I've heard?
<argylelabcoat>I see, I have to use `expose` for *each* serial port, and pass them into the /dev/ folder by name
<mdevos>oops, gcc is deprecated apparently, use gcc-toolchain
<mdevos>that said, I do not know if gcc is configured with Ada support by default
***Aurora_iz_kosmos is now known as Aurora_v_kosmose
<mdevos>daphnis: also, a quick ‘git grep -F ada’ doesn't find any packages using the Ada compiler
<mdevos>so it's quite possible guix support for Ada is minimal :-(
<daphnis>yeah, it doesn't look like there's support for it
<argylelabcoat>mdevos, specifically, I'm messing around with trying to create a container that can access a serial port. Ultimately, I want to be able to interface with USB serial device, but just as an experiment, I was doing this: `guix environment --ad-hoc python busybox -C --share=/var/lib/raptor=/var/lib/raptor --share=/dev/ttyACM0=/dev/ttyACM0`
<argylelabcoat>mdevos: seems like I can do them as an itemized list, but I was trying to figure out a way to make a deployable solution to access someone else's serial ports too... seems like the best solution to that right now is a VM image where each serial port can be bound up with qemu/virt-manager/whatever tool
<vertigo_38>Hi all! For testing purposes I just ran ```guix pack xterm``` and copied the resulting archive to my debian machine. I unpacked it, but when I try to execute the obviously existing xterm, I get a 'bash: gnu/store/751mcahyg3b5dpwkbfvzg6x1vdn9i49a-xterm-363/bin/xterm: file not found'. Does anyone have an idea how to get this working?
<argylelabcoat>my target arm board has ttymx* ports, which will be loads of fun, but I think to cross-compile a "container" to deploy to it with "guix copy" I'll have to manually enumerate the ports during the build
<mzbm>apteryx: thanks, so i have to install the firmware?
<mdevos>argylelabcoat: I just tried some random tty* from /dev/. Plenty are just virtual teletypes, aka terminal emulators
*mdevos don't know what /dev/ttyprintk is, though. Maybe it prints to to syslog or something?
<apteryx>so yes, without binary blobs the performance will be subpar. The Guixy thing to do would be get an older, second hand card which is known to be compatible (looking on h-node.org for example). That'll make your life easier and give you the most software freedom.
<apteryx>I myself use a pretty dated 8800 GTS card. Works well (no decoded video support but I don't watch 4K movies so I don't mind much)
<mzbm>apteryx: yeah i'll get another one if i can. nVidia is really not the right choice when going with linux. But AMD has no passively cooled cards. That was the only reason i've chosen the nvidia card.
<apteryx>mzbm: old nvidia is surprisingly superior than latest AMD, for all the publicity they get for their free drivers.
<apteryx>The catch is that their free drivers sometimes don't even boot without the proprietary binary blob installed.
<apteryx>That's based on dated experience (circa 2017?, amdgpu beginnigs IIRC) but I understand the situation hasn't improved much.
<lfam>Does the linter complain about use of git:// URLs?
<cybersyn>hi guix! i've been using guix as my primary os for the last 2.5 months and have to say it's brought back all of the excitement of my early computing experiences, and even surpassed it. but i haven't had too many questions so haven't had many reasons to interact with the community just yet.
<cybersyn>so here is a question that i figured is worth a gander: any cool packages that guix-specific worth checking out?
<cybersyn>wingo: thanks! I love watching your talks
<roptat>that kind of message is always nice to read, thanks :D
<roptat>not sure what you mean by guix-specific package?
<cybersyn>on another note, i should probably reintroduce myself: i'm the ux designer who was hanging out during guix days in november, and discussed some form of guix ux. at that time i was just starting on guix so wanted to give it some time, but i'm planning to write the mail list with thoughts after tet holiday. ultimately, its about researching through user feedback where there are pain points, and trying to understand how those problems
<cybersyn>could be adjusted for the community as a whole. not trying to "gui" guix, just understand some phenomenological problems
<cybersyn>ropat: i just meant if there are any packages that target guix specifically. i ask because i'm hoping to write a gui app that based on your shepherd-extensions dot file out put, gives you an interfaces for patching the diagram visually (like in pure data, for instance) and prints a guile config
<cybersyn>not sure if anyone has worked on such things
<roptat>we can generate a graphical representation of the service graph in d3.js (in a browser), but that's only for visualizing it, not modifying it
<cybersyn>yeah of course -- you don't js actually, you can use dot format
<cybersyn>but i didn't mean to inquire about the specifics of my project
<cybersyn>just to see if there are packages on guix that are designed specifically to enhance the guix experience. i'm thinking like melpa packages, but for guix
<cybersyn>awesome, guix-home-manager is something i'm unaware of :) and yes, gwl is rad
<cybersyn>theres a pet project thats been stirring i've been vaguely dabbling in on the weekends, which would be independent of the design effort, which would be a wayland modeline for navigating generations and the time machine, across different workspaces... think reproducible workspaces/desktops, along to navigate rebasing them, in a sense
<cybersyn>but thats something that would be independent of any ux effort, because in the end ux isn't about what a designer wants to do but instead reaching conclusions through research
<cybersyn>(i'm actually an artist, and only started ux since the pandemic, but enjoy the practicality/problem solving nature of it)
*wingo wonders why "guix time-machine" caused a guix update
<wingo>maybe i hadn't done it on this machine yet or something
<cybersyn>roptat: guix search guix-home-manager didn't return any results. perhaps its packaged under another name?
<roptat>it's not a package, it's a separate channel
<cybersyn>also folks, if ux isn't something the guix community would be interested in just let me know. i've been thinking through it and it raises a lot of difficult questions due to the novelty of question. but if the community is disinterested in a design research analysis of the project, i totally understand and dont want to impose
<cybersyn>it just seemed at guix days a quite a few were interested
***Alexander is now known as Guest27719
<civodul>cybersyn: i think a fresh analysis on it would be welcome!
<lfam>cybersyn: It's definitely an important topic for Guix. I wouldn't let a lack of response right now on IRC discourage you.
<Guest27719>Hello chat. I have a problem with the Guix gcc-toolchain 10.2.0 package. The list of default system include directories, printed with e.g. "g++ -xc++ -E -v /dev/null", does NOT contain /usr/include
<Guest27719>CMake probes the compiler for its default system include directories; trying to add those as an include directory with e.g. "include_directories(usr/include)" is normally silently ignored
<Guest27719>Since the Guix gcc-toolchain 10.2.0 compiler does not have /usr/include as a default include directory, however, it does NOT get ignored, and the resulting CMake build actually contains "-I/usr/include" command line parameters
<Guest27719>which obviously goes horribly wrong, as it mixes up header files from different gcc versions
<civodul>Guest27719: it's on purpose that Guix' gcc lacks /usr/include by default, because that directory "makes no sense" for Guix
<Guest27719>Is this a bug in the Guix gcc-toolchain 10.2.0 package? Shouldn't /usr/include be listed as a system include dir?
<civodul>in general, you shouldn't try to use Guix' gcc to build against libraries not provided by Guix
<Guest27719>On my own system, I have gcc 10.2.0 installed under /opt. It *does* have /usr/include compiled in as a system include dir
<cybersyn>civodul & lfam: hugs. i just want to contribute somehow because guix has enriched my life in so many ways and design seems to be the most impactful contribution i can offer. thanks for the encouragement
<civodul>lfam: BTW, i successfully upgraded my profile to "staging" and all is good so far